Mastering the game of Stratego with model-free multiagent reinforcement learning

https://doi.org/10.1126/science.add4679

DeepNash: Regularised Nash Dynamics (R-NaD), model-free, no search; top-3 2022 and all-time ranking on the Gravon Stratego platform against human experts.

Regret minimization and search for imperfect-information games